Improved photometric calibration of the SNLS and the SDSS supernova surveys

Context. We present a combined photometric calibration of the Supernova Legacy Survey (SNLS) and the SDSS supernova survey, which results from a joint effort of the SDSS and the SNLS collaborations. Aims. Our primary motivation is to eventually sharpen cosmological constraints derived from type Ia supernova measurements by improving the accuracy of the photometric calibration. We deliver fluxes calibrated to the HST spectrophotometric star network for large sets of tertiary stars that cover the science fields of both surveys in all photometric bands. We also cross-calibrate directly the two surveys and demonstrate their consistency. Methods. For each survey the flat-fielding is revised based on the analysis of dithered star observations. The calibration transfer from the HST spectrophotometric standard stars to the multi-epoch tertiary standard star catalogs in the science fields follows three different paths: observations of primary standard stars with the SDSS PT telescope; observations of Landolt secondary standard stars with SNLS MegaCam instrument at CFHT; and direct observation of faint HST standard stars with MegaCam. In addition, the tertiary stars for the two surveys are cross-calibrated using dedicated MegaCam observations of stripe 82. This overlap enables the comparison of these three calibration paths and justifies using their combination to improve the calibration accuracy. Results. Flat-field corrections have improved the uniformity of each survey as demonstrated by the comparison of photometry in overlapping fields: the rms of the difference between the two surveys is 3 mmag in gri, 4 mmag in z and 8 mmag in u. we also find a remarkable agreement (better than 1%) between the SDSS and the SNLS calibration in griz. The cross-calibration and the introduction of direct calibration observations bring redundancy and strengthen the confidence in the resulting calibration. We conclude that the surveys are calibrated to the HST with precision of about 0.4% in griz. This precision is comparable to the external uncertianty affecting the color of the HST primary standard stars.